Pith. sign in

REVIEW 4 major objections 5 minor 84 references

DataWink: Reusing and Adapting SVG-based Visualization Examples with Large Multimodal Models

T0 review · 4 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read DataWink claims that a chain of large multimodal models can turn an existing SVG-based visualization—even one with heavy decoration and non-standard encodings—into a reusable, parameterized template, so that non-experts can swap in new…

desk verdict A genuinely useful systems paper that turns decorated SVG charts into adaptable templates via an LMM pipeline, with an honest user study whose evidence is narrower than the conclusion claims. read the letter →

arxiv 2507.17734 v1 pith:4F4MQ4AI submitted 2025-07-23 cs.HC

Liwenhan Xie , Yanna Lin , Can Liu , Huamin Qu , Xinhuan Shu This is my paper · ORCID
classification cs.HC
keywords visualizationbyexamplelargemultimodalmodelsSVGreusetemplatesintermediaterepresentationlazydatabindingdynamicwidgetsuserstudy
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

DataWink claims that a chain of large multimodal models can turn an existing SVG-based visualization—even one with heavy graphical decoration and non-standard encodings—into a reusable template that accepts new data and editable style parameters. The paper's central bet is that this removes the hardest part of by-example authoring: figuring out how the picture maps to the data. A user study with twelve participants found that people could adapt a decorated bar chart to new data and redesign it more easily, and rated their replicated outputs higher, than when using familiar commercial tools. If the pipeline holds up, non-designers can inherit the aesthetic quality of professionally designed charts instead of rebuilding them from scratch.

What carries the argument

The load-bearing object is the intermediate representation of a visualization: a layered abstraction that splits an SVG into data-driven layers (marks, axes, legends), text layers, decorative layers, and configuration layers, with per-layer properties such as mark types, data-encoded attributes, origin, and bounding box. It is produced by a three-step LMM chain—role identification and data extraction, semantic enrichment with natural-language descriptions, and final representation generation—and it is what lets the pipeline regenerate a D3 template rather than merely re-skin the original file. The second mechanism is dynamic refinement: user requests are converted into minimal edits to the template and into on-demand widgets (sliders, color pickers, inputs), whose parameters are linked back to the program, so fine-tuning happens through direct manipulation rather than repeated prompting.

What would settle it

Give the pipeline a set of decorated SVG charts with known ground-truth data and check whether the recovered numbers match the originals and whether, after swapping in new data, decorative dependencies such as shadow lengths and contact points still track their parent marks; a substantial failure rate on either check would refute the central claim.

Watch

Extended reading notes

Core claim

The paper proposes a two-phase pipeline. In the decomposition phase, an SVG reference is preprocessed—each visual element gets an identity and redundant markup is stripped—and then a chained LMM reads the simplified SVG plus a downscaled raster rendering to group elements into four layers: data-driven, text, decorative, and configuration. The chain also extracts the underlying dataset and enriches element groups with natural-language descriptions, producing an intermediate representation that stands between raw SVG and a visualization program. In the construction phase, another LMM synthesizes a D3 program that regenerates the data-driven layers from parameters, preserving the original design including dependencies such as shadow shapes tied to parent bars. The DataWink interface wraps this template in a data table, a template inspector, direct canvas manipulation, and a chat agent that turns user requests into live widgets. The paper's claim is that this combination makes reuse and adaptation faster and more effective than conventional editing or generic LMM prompting.

Load-bearing premise

The load-bearing premise is that the large multimodal model chain can reliably read the structure and data values out of a simplified SVG picture and write an editable template that keeps the decorative relationships intact; if that works only for simple or regular charts, the central claim about general reuse and democratization is not established.

Editorial extensions

If this is right

  • A user with no D3 experience can load a decorated SVG, upload a CSV with the right column types, and get a chart that keeps the reference's graphical design, because data binding is handled by the generated template.
  • Adaptation requests expressed in natural language do not require the user to know SVG internals: the system creates widgets for new parameters, letting the user steer values directly.
  • Non-standard decorative dependencies—such as shadow parallelograms whose angles and contact points must track their parent bars—can survive a data swap, which the user study showed is the step where conventional tools fail.
  • Because the template exposes the intermediate representation and generated program, advanced users can inspect or patch the mapping, making the AI's understanding auditable rather than a black box.
  • By-example authoring of this kind can lower the barrier to personalized, aesthetically rich visualization, since the 'design' step is reduced to choosing and adapting an existing high-quality example.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A testable extension is whether the same role-labeled intermediate representation could also feed accessibility tools, chart search, or automated style transfer, since the paper already makes those roles explicit.
  • The pipeline's reliability is likely to degrade as SVG structure becomes less regular; the paper itself sets hand-drawn sketches and bitmap infographics out of scope, so a stress test on icon-heavy or pictorial charts with shared masks would reveal how much of the claim generalizes.
  • Because the paper identifies propagated errors in early LMM chain steps as a failure mode, a concrete improvement left implicit is a verification step in which the model compares its generated SVG against the reference before presenting a template.
  • The published prompt templates and generated intermediate representations could double as a benchmark suite for measuring how well LMMs understand SVG structure, giving the visualization community a reusable testbed.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper presents DataWink, an LMM-powered pipeline and interface for reusing and adapting SVG-based visualization examples. The pipeline decomposes an input SVG into an intermediate representation (data-driven, text, decorative, and configuration layers), then synthesizes a D3-based template with parameterized controls; the interface supports data replacement, chat-based adaptation, dynamically generated widgets, and direct manipulation. The evaluation is a within-subjects user study (N=12) with a replication task comparing DataWink against a mix of familiar commercial tools, followed by an open-ended redesign task and Likert questionnaires. The authors claim that DataWink enables faster and more effective reuse and adaptation of visualizations compared with conventional tools while aligning with users' creative workflows.

Significance. If the broad claim holds, DataWink would be a valuable contribution to example-driven visualization authoring, extending prior work on standard charts and timelines to decorated, non-standard SVG visualizations. The paper's strengths include a well-motivated intermediate representation, a transparently reported and reproducible implementation (source code and prompt templates are released), a structured user study with qualitative findings, and an unusually candid limitations section. However, the central generalization is not yet fully established: the quantitative user-study comparison rests on a single curated reference chart, and Section 7.1 explicitly concedes that the LMM chain is prone to propagated errors that constrain applicability to more advanced visualization types. The contribution is promising and publishable in principle, but the evidence as presented supports a more narrowly scoped claim than the conclusion states.

major comments (4)
  1. [§6.4, Fig. 5] The sentence stating that 'participants in the DataWink condition achieved a significantly higher performance' is not supported by any reported statistical test for the completion-rate data. Fig. 5 shows only percentages, with no p-values, effect sizes, or per-subtask raw counts. Please either add an appropriate statistical analysis (e.g., McNemar or a mixed-effects model on subtask completion) or remove the word 'significantly' and describe the result as a descriptive difference.
  2. [§6.1, §6.4, §7.1] The quantitative comparison in Task A uses a single reference visualization (the Window chart), and the gallery adds only three author-selected examples. Section 7.1 concedes that the method 'operates upon LMM chains, which are prone to propagated errors at early stages' and that this 'constrains its applicability to more advanced visualization types.' Since all downstream user-facing capabilities presuppose reliable decomposition and templating, the conclusion that DataWink enables faster and more effective reuse and adaptation of visualizations in general is not established by the current evidence. I suggest either adding a pipeline-reliability evaluation over a broader corpus of decorated SVG charts or carefully qualifying the conclusion to the demonstrated chart class.
  3. [§6.1] The baseline condition is not a fixed tool: participants could mix any commercial tools they were familiar with, including ChatGPT and Gemini, which are the same model family that powers DataWink. This confound makes it difficult to attribute the observed differences to the DataWink pipeline and interface rather than to tool choice, prior familiarity, or the absence of a structured pipeline in the baseline. Please report which tools each participant actually used and discuss whether the comparison holds when controlling for tool familiarity or model family.
  4. [§6.4, Fig. 5] There is a potential inconsistency between the claim that all N=12 participants 'could finish data adaptation tasks, maintain original visual decorations, and perform global chart adjustment' and the statement that N=4 participants 'failed to control the data-driven shadows and ran out of time before synthesizing accurate gradients.' Please clarify which subtasks count toward completion, how the completion criteria were applied uniformly, and how partial completion or timeout was coded in the reported rates.
minor comments (5)
  1. [§6.4, Fig. 5] The completion-rate visualization should include raw counts or confidence intervals, and the figure caption should state the denominator for each percentage to make the rates interpretable with N=12.
  2. [§3.2] The intermediate representation is described with a bullet-list pseudo-grammar; a formal schema or a small JSON example would make the representation concrete and facilitate replication by other researchers.
  3. [§4.4] The GitHub URL is broken across a line break in the rendered text; please provide a single clickable URL for the source code and prompts.
  4. [§6.2] The limitation paragraph in §6.1 mentions sample size and baseline tool choice, but not the participants' shared cultural and linguistic background, which is relevant to the generalizability of findings about natural-language interaction and workflow fit.
  5. [Fig. 6] Since reverse-coded items were flipped for reporting, the caption should indicate the original wording of those items or note that all ratings were normalized so that higher values imply stronger agreement.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the central claims rest on a user study and a concrete SVG-to-template pipeline, not on fitted inputs or self-citation chains.

full rationale

The paper's derivation chain is not circular. The core technical claim is that an LMM pipeline can convert an SVG visualization into an intermediate representation and then into a D3-based template; the core evaluation claim is that users complete replication and redesign tasks faster and more effectively with DataWink than with familiar commercial tools. Neither claim is defined in terms of the other. The intermediate representation is defined independently of the template output (Section 3.2), and the pipeline's outputs are produced by LMM prompts on a preprocessed SVG plus a raster image, not by fitting parameters to the evaluation results. The template synthesis step (Section 3.4.1) uses the parsed original data to replicate the reference, but this is an implementation detail of the system under test, not a hidden reuse of the evaluation outcome. The user study (Section 6) is an empirical, externally observable comparison with a counterbalanced within-subjects design, screen recordings, task completion rates, and questionnaire responses; the completion criteria in Fig. 5 are concrete output states rather than LMM self-assessments or the paper's own predicted scores. The paper honestly discloses load-bearing limitations in Section 7.1, including LMM-chain error propagation, long prompts, and restricted applicability to more advanced visualization types; this transparency does not make the method circular. There are self-citations to prior works by the same authors (e.g., MetaGlyph [77], WaitGPT [75], Wakey-Wakey [76], GVVST [60], and the timeline work [84]), but these are used as related-work context and inspiration, not as the justification for DataWink's central claims; no uniqueness theorem or load-bearing result is imported from them. The code and prompt templates are made publicly available, providing independent reproducibility support. A weak external ground truth for extracted data and template fidelity is a validity concern, not a circularity concern, because the paper never asserts an equation or a fitted quantity that is equivalent to its own input by construction.

Assumptions & free parameters 4 free parameters · 4 assumptions · 3 invented entities

The central claim rests on three categories of assumptions: the fidelity of SVG as a source for recovering data encoding, the capability of GPT-based models to perform the decomposition and synthesis reliably, and the representativeness of the user study. None are proven by the paper itself; the LMM capability is explicitly acknowledged as limited in Section 7.1. The design choices listed as free parameters were selected by hand and are not derived from a theory.

free parameters (4)
  • SVG decimal precision threshold = 2 decimal places
    Used in preprocessing to reduce control point precision; chosen by hand to balance token budget and appearance, but not derived from any requirement.
  • Raster image max width = 400 pixels
    Empirically set for the LMM's visual context in Phase I; affects text legibility and token count.
  • One-shot example chart = A bar chart example
    Used in prompts for role identification and template synthesis; performance may depend on the choice of this illustrative example.
  • LMM backbone = GPT-4o-mini, GPT-4o-128k
    Hand-selected models via OpenAI API; the system's behavior is tied to these proprietary model versions and their prompt sensitivity.
assumptions (4)
  • domain assumption SVG files contain enough structural information to recover the data values and encoding scheme of the original visualization.
    The pipeline takes any SVG as input and assumes coordinates, paths, and marks map deterministically to underlying data. The paper does not prove this holds for arbitrary creative designs; free-form and masked elements can obscure the mapping.
  • ad hoc to paper Large multimodal models (GPT-4o and GPT-4o-mini) can reliably perform role identification, data extraction, and D3 code synthesis from a simplified SVG plus a raster image.
    The entire pipeline depends on this; Section 7.1 concedes the models have limited SVG understanding and the chain propagates early errors.
  • domain assumption The proposed four-layer intermediate representation is expressive enough to cover the target class of decorated, data-driven SVG visualizations.
    No formal coverage argument is given; the paper demonstrates three gallery examples and the user study's single reference chart, which is not strong evidence of generality.
  • domain assumption User study participants' behavior and self-reports predict the experience of the intended non-expert user population.
    Most participants had years of design or visualization experience, so the democratizing claim rests on an experienced sample rather than novices.
invented entities (3)
  • Intermediate representation of visualizations independent evidence
    purpose: Four-layer abstraction (data-driven, text, decorative, configuration) used to bridge primitive SVG elements to D3 template synthesis.
    The layer structure is defined by the authors; its adequacy is testable through the gallery and user study, and the code is available. It does not make an external falsifiable prediction beyond system behavior.
  • Slot-based SVG (marked-up SVG with custom XML markers) independent evidence
    purpose: Enables LMMs to group and describe SVG elements without breaking existing styles during template generation.
    The markers are non-standard SVG; their utility is evidenced only through the three gallery examples and user study, not a formal analysis. It is testable by running the pipeline on new SVGs.
  • Dynamic widget synthesis independent evidence
    purpose: Generates sliders, color pickers, and inputs from natural-language adaptation requests.
    Usage is quantified in the user study (average 8.50 widgets per participant), and the behavior is observable in the shipped system. No separate benchmark validates widget correctness.

how reviews work

0 comments
Cite this review

Pith. "Pith review of DataWink: Reusing and Adapting SVG-based Visualization Examples with Large Multimodal Models." pith.science (2026). https://pith.science/paper/4F4MQ4AI

@misc{pith2026250717734,
  author       = {Pith},
  title        = {Pith review of: DataWink: Reusing and Adapting SVG-based Visualization Examples with Large Multimodal Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4F4MQ4AI}},
  note         = {Machine review of arXiv:2507.17734}
}
read the original abstract

Creating aesthetically pleasing data visualizations remains challenging for users without design expertise or familiarity with visualization tools. To address this gap, we present DataWink, a system that enables users to create custom visualizations by adapting high-quality examples. Our approach combines large multimodal models (LMMs) to extract data encoding from existing SVG-based visualization examples, featuring an intermediate representation of visualizations that bridges primitive SVG and visualization programs. Users may express adaptation goals to a conversational agent and control the visual appearance through widgets generated on demand. With an interactive interface, users can modify both data mappings and visual design elements while maintaining the original visualization's aesthetic quality. To evaluate DataWink, we conduct a user study (N=12) with replication and free-form exploration tasks. As a result, DataWink is recognized for its learnability and effectiveness in personalized authoring tasks. Our results demonstrate the potential of example-driven approaches for democratizing visualization creation.

Figures

Figures reproduced from arXiv: 2507.17734 by the authors.

Figure 1
Figure 1. We introduce DataWink, an interactive by-example authoring tool for reusing and adapting SVG-based visualizations. Powered [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. An overview of the pipeline. (I) Decomposing an example visualization into an intermediate representation. (II) Constructing reusable [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. A screenshot of DataWink, the proof-of-concept authoring interface for reusing and adapting SVG-based examples. (A) Canvas panel–Users [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: A gallery of visualizations made with DataWink reusing reference SVGs (top left) and adapting the visual design. [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: An overview of tasks in the user evaluation. In Task A, participants are provided with a reference SVG, a dataset, and three images to [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: Distribution of user ratings based on a 7-point Likert scale. 1: [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

84 extracted references · 18 canonical work pages

  1. [3]

    Bigelow, S

    A. Bigelow, S. Drucker, D. Fisher, and M. Meyer. Reflections on how designers design with data. In Proc. AVI, pp. 17–24. ACM, 2014. doi: 10. 1145/2598153.2598175 2

    arXiv 2014

  2. [5]

    M. A. Borkin, Z. Bylinskii, N. W. Kim, C. M. Bainbridge, C. S. Yeh, D. Borkin, H. Pfister, and A. Oliva. Beyond memorability: Visualization recognition and recall. IEEE Trans. Vis. Comput. Graph., 22(1):519–528,

  3. [6]

    Bostock, V

    M. Bostock, V . Ogievetsky, and J. Heer. D 3: Data-driven documents. IEEE Trans. Vis. Comput. Graph., 17(12):2301–2309, 2011. doi: 10.1109/ TVCG.2011.185 3, 4, 9

    2011

  4. [7]

    A. Cairo. The art of insight: How great visualization designers think. John Wiley & Sons, 2023. 1

    2023

  5. [8]

    C. Chen, H. K. Bako, P. Yu, J. Hooker, J. Joyal, S. C. Wang, S. Kim, J. Wu, A. Ding, L. Sandeep, A. Chen, C. Sinha, and Z. Liu. VisAnatomy: An SVG chart corpus with fine-grained semantic labels. arXiv:2410.12268,

    work page Pith review arXiv

  6. [9]

    C. Chen, B. Lee, Y . Wang, Y . Chang, and Z. Liu. Mystique: Deconstructing SVG charts for layout reuse. IEEE Trans. Vis. Comput. Graph., 30(1):447– 457, 2024. doi: 10.1109/TVCG.2023.3327354 2

    arXiv 2024

  7. [10]

    Q. Chen, W. Shuai, J. Zhang, Z. Sun, and N. Cao. Beyond numbers: Creating analogies to enhance data comprehension and communication with generative AI. In Proc. CHI, article no. 377, 14 pages. ACM, 2024. doi: 10.1145/3613904.3642480 2

    arXiv 2024

  8. [11]

    Cheng, T

    R. Cheng, T. Barik, A. Leung, F. Hohman, and J. Nichols. BISCUIT: Scaffolding LLM-generated code with ephemeral UIs in computational notebooks. In Proc. VL/HCC, pp. 13–23. IEEE, 2024. doi: 10.1109/VL/ HCC60511.2024.00012 2

    arXiv 2024

Show all 84 references
  1. [12]

    Coelho and K

    D. Coelho and K. Mueller. Infomages: Embedding data into thematic images. Comput. Graph. Forum, 39(3):593–606, 2020. doi: 10.1111/cgf. 14004 1

    2020 doi

  2. [14]

    Dahary, M

    O. Dahary, M. Lu, O. Patashnik, and D. Cohen-Or. PictorialAttributes: Depicting multiple attributes with realistic imaging. In ACM SIGGRAPH 2024 Posters, article no. 69, 2 pages. ACM, 2024. doi: 10.1145/3641234. 3671072 9

    2024 doi

  3. [15]

    Di Bartolomeo, V

    S. Di Bartolomeo, V . Schetinger, J. L. Adams, A. M. McNutt, M. El- Assady, and M. Miller. Doom or deliciousness: Challenges and opportu- nities for visualization in the age of generative models. Comput. Graph. Forum, 42(3):423–435, 2023. doi: 10.1111/cgf.14841 2

    2023 doi

  4. [16]

    S. Dou, X. Jiang, L. Liu, L. Ying, C. Shan, Y . Shen, X. Dong, Y . Wang, D. Li, and C. Zhao. Hierarchically recognizing vector graphics and a new chart-based vector graphics dataset. IEEE Trans. Pattern Anal. Mach. Intell., 46(12):7556–7573, 2024. doi: 10.1109/TPAMI.2024.3394298 2

    2024

  5. [17]

    J. Ge, Z. Z. Wang, X. Zhou, Y .-H. Peng, S. Subramanian, Q. Tan, M. Sap, A. Suhr, D. Fried, G. Neubig, and T. Darrell. AutoPresent: Designing structured visuals from scratch. In Proc. CVPR, 2025. arXiv:2501.00912. 9

    2025 arXiv

  6. [18]

    Goswami, P

    K. Goswami, P. Mathur, R. Rossi, and F. Dernoncourt. PlotEdit: Natural language-driven accessible chart editing in PDFs via multimodal LLM agents. In Proc. ECIR, pp. 130–134. Springer, 2025. doi: 10.1007/978-3 -031-88720-8_22 1

    2025 doi

  7. [19]

    Harper and M

    J. Harper and M. Agrawala. Deconstructing and restyling D3 visualiza- tions. In Proc. UIST, pp. 253–262. ACM, 2014. doi: 10.1145/2642918. 2647411 2

    2014 doi

  8. [20]

    Harper and M

    J. Harper and M. Agrawala. Converting basic D3 charts into reusable style templates. IEEE Trans. Vis. Comput. Graph., 24(3):1274–1286, 2018. doi: 10.1109/TVCG.2017.2659744 2

    2018

  9. [21]

    Harrison, K

    L. Harrison, K. Reinecke, and R. Chang. Infographic aesthetics: Designing for the first impression. In Proc. CHI, pp. 1187–1190. ACM, 2015. doi: 10.1145/2702123.2702545 1

    2015

  10. [22]

    Y . He, K. Xu, S. Cao, Y . Shi, Q. Chen, and N. Cao. Leveraging foundation models for crafting narrative visualization: A survey. IEEE Trans. Vis. Comput. Graph., 2025. Early Access. doi: 10.1109/TVCG.2025.3542504 2

    2025

  11. [23]

    D. Jung, W. Kim, H. Song, J.-i. Hwang, B. Lee, B. Kim, and J. Seo. ChartSense: Interactive data extraction from chart images. In Proc. CHI, pp. 6706–6717. ACM, 2017. doi: 10.1145/3025453.3025957 2

    2017

  12. [24]

    Kembhavi, M

    A. Kembhavi, M. Salvato, E. Kolve, M. Seo, H. Hajishirzi, and A. Farhadi. A diagram is worth a dozen images. InProc. ECCV, pp. 235–251. Springer,

  13. [25]

    N. W. Kim, E. Schweickart, Z. Liu, M. Dontcheva, W. Li, J. Popovic, and H. Pfister. Data-Driven Guides: Supporting expressive design for information graphics. IEEE Trans. Vis. Comput. Graph., 23(1):491–500,

  14. [26]

    doi: 10.1007/978-3-319-46493-0_15 2

    doi

  15. [27]

    Kumar, J

    R. Kumar, J. O. Talton, S. Ahmad, and S. R. Klemmer. Bricolage: Example-based retargeting for web design. In Proc. CHI, pp. 2197–2206. ACM, 2011. doi: 10.1145/1978942.1979262 2

    2011

  16. [28]

    doi: 10.1109/TVCG.2016.2598620 2

    2016

  17. [29]

    T. S. Kim, D. Choi, Y . Choi, and J. Kim. Stylette: Styling the web with natural language. In Proc. CHI, article no. 5, 17 pages. ACM, 2022. doi: 10.1145/3491102.3501931 1

    2022

  18. [30]

    V . Liu, R. H. Kazi, L.-Y . Wei, M. Fisher, T. Langlois, S. Walker, and L. Chilton. Logomotion: Visually-grounded code synthesis for creating and editing animation. In Proc. CHI, article no. 157, 16 pages. ACM,

  19. [31]

    J. R. Lewis. The system usability scale: Past, present, and future. Int. J. Hum.-Comput. Interact., 34(7):577–590, 2018. doi: 10.1080/10447318. 2018.1455307 7

    2018

  20. [32]

    T. F. Liu, M. Craft, J. Situ, E. Yumer, R. Mech, and R. Kumar. Learning design semantics for mobile apps. In Proc. UIST, 11 pages, pp. 569–579. ACM, 2018. doi: 10.1145/3242587.3242650 2

    2018

  21. [33]

    M. Lu, C. Wang, J. Lanir, N. Zhao, H. Pfister, D. Cohen-Or, and H. Huang. Exploring visual information flows in infographics. In Proc. CHI, article no. 136, 12 pages. ACM, 2020. doi: 10.1145/3313831.3376263 2

    2020

  22. [34]

    Y . Lu, A. Leung, A. Swearngin, J. Nichols, and T. Barik. Misty: UI prototyping through interactive conceptual blending. In Proc. CHI, article no. 1108, 17 pages. ACM, 2025. doi: 10.1145/3706598.3713924 2

    2025

  23. [36]

    Z. Liu, J. Thompson, A. Wilson, M. Dontcheva, J. Delorey, S. Grigg, B. Kerr, and J. Stasko. Data Illustrator: Augmenting vector design tools with lazy data binding for expressive visualization authoring. In Proc. CHI, 13 pages. ACM, 2018. doi: 10.1145/3173574.3173697 2

    2018

  24. [37]

    Madan, Z

    S. Madan, Z. Bylinskii, C. Nobre, M. Tancik, A. Recasens, K. Zhong, S. Alsheikh, A. Oliva, F. Durand, and H. Pfister. Parsing and summarizing infographics with synthetically trained icon detection. In Proc. PacificVis, pp. 31–40, 2021. doi: 10.1109/PacificVis52677.2021.00012 2

    2021

  25. [38]

    Masson, S

    D. Masson, S. Malacria, D. V ogel, E. Lank, and G. Casiez. ChartDetective: Easy and accurate interactive data extraction from complex vector charts. In Proc. CHI, article no. 147, 17 pages. ACM, 2023. doi: 10.1145/3544548 .3581113 2

    2023 doi

  26. [39]

    J. Luo, Z. Li, J. Wang, and C.-Y . Lin. ChartOCR: Data extraction from charts images via a deep hybrid framework. In Proc. WACV, pp. 1917– 1925, 2021. doi: 10.1109/W ACV48630.2021.00196 2

    1917

  27. [40]

    G. Lupi, S. Posavec, and M. Popova. Dear Data. Princeton Architectural Press, 2016. 1

    2016

  28. [41]

    Offenwanger, T

    A. Offenwanger, T. Tsandilas, and F. Chevalier. DataGarden: Formalizing personal sketches into structured visualization templates. IEEE Trans. Vis. Comput. Graph., 31(1):1268–1278, 2025. doi: 10.1109/TVCG.2024. 3456336 2, 3, 9

    2025 doi

  29. [42]

    P. Parsons. Understanding data visualization design practice. IEEE Trans. Vis. Comput. Graph., 28(1):665–675, 2023. doi: 10.1109/TVCG.2021. 3114959 1

    2023 doi

  30. [43]

    A. M. McNutt and R. Chugh. Integrated visualization editing via parame- terized declarative templates. In Proc. CHI, article no. 17, 14 pages. ACM,

  31. [44]

    Pollock, C

    J. Pollock, C. Mei, G. Huang, E. Evans, D. Jackson, and A. Satyanarayan. Bluefish: Composing diagrams with declarative relations. In Proc. UIST, article no. 23, 21 pages, 2024. doi: 10.1145/3654777.3676465 9

    2024

  32. [45]

    G. G. Méndez, M. A. Nacenta, and S. Vandenheste. iV oLVER: Interactive visual language for visualization extraction and reconstruction. In Proc. CHI, pp. 4073–4085. ACM, 2016. doi: 10.1145/2858036.2858435 2

    2016

  33. [47]

    D. Ren, B. Lee, and M. Brehmer. Charticulator: Interactive construction of bespoke chart layouts. IEEE Trans. Vis. Comput. Graph., 25(1):789–799,

  34. [48]

    Poco and J

    J. Poco and J. Heer. Reverse-engineering visualizations: Recovering visual encodings from chart images. Comput. Graph. Forum, 36(3):353–363,

  35. [49]

    Satyanarayan, D

    A. Satyanarayan, D. Moritz, K. Wongsuphasawat, and J. Heer. Vega-lite: A grammar of interactive graphics. IEEE Trans. Vis. Comput. Graph., 23(1):341–350, 2016. doi: 10.1109/TVCG.2016.2599030 9

    2016

  36. [50]

    Savva, N

    M. Savva, N. Kong, A. Chhajta, L. Fei-Fei, M. Agrawala, and J. Heer. ReVision: Automated classification, analysis and redesign of chart images. In Proc. UIST, pp. 393–402, 2011. doi: 10.1145/2047196.2047247 2

    2011

  37. [51]

    A. Qi, T. Tsandilas, A. Shamir, and A. Bousseau. Sketch2Data: Recov- ering data from hand-drawn infographics. Comput. Graph., article no. 104251, 14 pages, 2025. doi: 10.1016/j.cag.2025.104251 2, 9

    2025

  38. [52]

    L. Shen, H. Li, Y . Wang, T. Luo, Y . Luo, and H. Qu. Data Playwright: Authoring data videos with annotated narration. IEEE Trans. Vis. Comput. Graph., 2024. Early Access. doi: 10.1109/TVCG.2024.3477926 2

    2024

  39. [53]

    C. Shi, C. Yang, Y . Liu, B. Shui, J. Wang, M. Jing, L. Xu, X. Zhu, S. Li, Y . Zhang, G. Liu, X. Nie, D. Cai, and Y . Yang. ChartMimic: Evaluating LMM’s cross-modal reasoning capability via chart-to-code generation. In Proc. ICLR, 57 pages, 2025. arXiv:2406.09961. 1, 2

    2025 arXiv

  40. [54]

    X. Shi, Y . Wang, R. Rossi, and J. Zhao. Brickify: Enabling expressive design intent specification through direct manipulation on design tokens. In Proc. CHI, article no. 424, 20 pages. ACM, 2025. doi: 10.1145/3706598 .3714087 9

    2025 doi

  41. [55]

    Satyanarayan, B

    A. Satyanarayan, B. Lee, D. Ren, J. Heer, J. Stasko, J. Thompson, M. Brehmer, and Z. Liu. Critical reflections on visualization author- ing systems. IEEE Trans. Vis. Comput. Graph., 26(1):461–471, 2019. doi: 10.1109/TVCG.2019.2934281 2

    2019

  42. [57]

    H. V . Shin, J. Warner, B. Hartmann, C. Gomes, H. Winnemoeller, and W. Li. Multi-level correspondence via graph kernels for editing vector graphics designs. In Proc. GI, pp. 97–107. CHCSS/SCDHM, 2021. doi: 10.20380/GI2021.12 2

    2021 doi

  43. [58]

    Schöttler, J

    S. Schöttler, J. Dykes, J. Wood, U. Hinrichs, and B. Bach. Constraint-based breakpoints for responsive visualization design and development. IEEE Trans. Vis. Comput. Graph., 2024. Early Access. doi: 10.1109/TVCG. 2024.3410097 9

    2024

  44. [59]

    D. V . Society. Information is beautiful awards. https://www. informationisbeautifulawards.com/about, 2025. Accessed on March 25, 2025. 9

    2025

  45. [60]

    S. Song, Y . Zhang, Y . Lin, H. Qu, C. Wang, and C. Li. GVVST: Image- driven style extraction from graph visualizations for visual style transfer. IEEE Trans. Vis. Comput. Graph. , 2024. Early Access. doi: 10.1109/ TVCG.2024.3485701 2

    2024

  46. [61]

    Stefaner

    M. Stefaner. Crisis? What crisis?—Why 2024 was a dead year for indie dataviz – and how we’ll do much better in 2025. https: //truth-and-beauty.net/texts/crisis-what-crisis . Accessed on March 25, 2025. 9

    2024

  47. [62]

    Y . Shi, T. Gao, X. Jiao, and N. Cao. Understanding design collaboration between designers and artificial intelligence: A systematic literature review. Proceedings of the ACM on Human-Computer Interaction, 7(CSCW2), article no. 368, 35 pages, 2023. doi: 10.1145/3610217 2

    2023 doi

  48. [63]

    Tsandilas

    T. Tsandilas. StructGraphics: Flexible visualization design through data- agnostic and reusable graphical structures. IEEE Trans. Vis. Comput. Graph., 27(2):315–325, 2020. doi: 10.1109/TVCG.2020.3030476 2

    2020

  49. [64]

    Vaithilingam, E

    P. Vaithilingam, E. L. Glassman, J. P. Inala, and C. Wang. DynaVis: Dy- namically synthesized UI widgets for visualization editing. In Proc. CHI, article no. 985, 17 pages. ACM, 2024. doi: 10.1145/3613904.3642639 1, 2, 4

    2024

  50. [65]

    L. S. Snyder, C. Wang, and S. Drucker. Challenges & opportunities with llm-assisted visualization retargeting. ArXiv Preprint, 5 pages, 2025. doi: 10.48550/arXiv.2507.01436 1

    2025 doi

  51. [66]

    Warner, K

    J. Warner, K. W. Kim, and B. Hartmann. Interactive flexible style transfer for vector graphics. In Proc. UIST, article no. 49, 14 pages. ACM, 2023. doi: 10.1145/3586183.3606751 2

    2023

  52. [67]

    Wilkinson

    L. Wilkinson. The grammar of graphics. In Handbook of computational statistics: Concepts and methods, pp. 375–414. Springer, 2011. doi: 10. 1007/0-387-28695-0 9

    2011

  53. [68]

    A. Wu, Y . Wang, X. Shu, D. Moritz, W. Cui, H. Zhang, D. Zhang, and H. Qu. AI4VIS: Survey on artificial intelligence approaches for data visualization. IEEE Trans. Vis. Comput. Graph., 28(12):5049–5070, 2022. doi: 10.1109/TVCG.2021.3099002 2

    2022

  54. [69]

    Y . Tian, W. Cui, D. Deng, X. Yi, Y . Yang, H. Zhang, and Y . Wu. ChartGPT: Leveraging LLMs to generate charts from abstract natural language. IEEE Trans. Vis. Comput. Graph., 31(3):1731–1745, 2025. doi: 10.1109/TVCG. 2024.3368621 2

    2025

  55. [70]

    S. Wu. What killed innovation? https://www.shirleywu.studio/ notebook/2025-02-innovation-killer . Accessed on March 25,

    2025

  56. [71]

    H. Xia, N. Henry Riche, F. Chevalier, B. De Araujo, and D. Wigdor. DataInk: Direct and creative data-oriented drawing. In Proc. CHI, article no. 223, 13 pages. ACM, 2018. doi: 10.1145/3173574.3173797 3

    2018

  57. [72]

    C. Wang, B. Lee, S. Drucker, D. Marshall, and J. Gao. Data Formulator 2: iterative creation of data visualizations, with AI transforming data along the way. In Proc. CHI, article no. 677, 17 pages. ACM, 2025. doi: 10. 1145/3706598.3713296 2

    2025

  58. [73]

    S. Xiao, S. Huang, Y . Lin, Y . Ye, and W. Zeng. Let the chart spark: Embedding semantic context into chart with text-to-image generative model. IEEE Trans. Vis. Comput. Graph., 30(1):284–294, 2023. doi: 10. 1109/TVCG.2023.3326913 2, 9

    2023

  59. [74]

    S. Xiao, L. Wang, X. Ma, and W. Zeng. TypeDance: Creating seman- tic typographic logos from image through personalized generation. In Proc. CHI, article no. 175, 18 pages. ACM, 2024. doi: 10.1145/3613904. 3642185 2

    2024 doi

  60. [75]

    L. Xie, C. Zheng, H. Xia, H. Qu, and C. Zhu-Tian. WaitGPT: Monitoring and steering conversational LLM agent in data analysis with on-the-fly code visualization. In Proc. UIST, article no. 52, 14 pages. ACM, 2024. doi: 10.1145/3654777.3676374 2

    2024

  61. [76]

    R. Wu, W. Su, and J. Liao. Chat2SVG: Vector graphics generation with large language models and image diffusion models. In Proc. CVPR, 2025. arXiv:2411.16602. 9

    2025

  62. [77]

    L. Ying, X. Shu, D. Deng, Y . Yang, T. Tang, L. Yu, and Y . Wu. MetaGlyph: Automatic generation of metaphoric glyph-based visualization. IEEE Trans. Vis. Comput. Graph., 29(1):331–341, 2022. doi: 10.1109/TVCG. 2022.3209447 1, 2

    2022

  63. [78]

    L.-P. Yuan, W. Zeng, S. Fu, Z. Zeng, H. Li, C.-W. Fu, and H. Qu. Deep colormap extraction from visualizations. IEEE Trans. Vis. Comput. Graph., 28(12):4048–4060, 2021. doi: 10.1109/TVCG.2021.3070876 2

    2021

  64. [79]

    S. Xiao, Y . Hou, C. Jin, and W. Zeng. WYTIWYR: A user intent-aware framework with multi-modal inputs for visualization retrieval. Comput. Graph. Forum, 42(3):311–322, 2023. doi: 10.1111/cgf.14832 9

    2023 doi

  65. [80]

    Zhang, J

    S. Zhang, J. Ma, J. Wu, D. Ritchie, and M. Agrawala. Editing motion graphics video via motion vectorization and transformation. ACM Trans. Gr., 42(6), article no. 229, 13 pages, 2023. doi: 10.1145/3618316 2

    2023 doi

  66. [81]

    Zhang, R

    Y . Zhang, R. Jiang, L. Xie, Y . Zhao, C. Liu, T. Ding, S. Chen, and X. Yuan. OldVisOnline: Curating a dataset of historical visualizations. IEEE Trans. Vis. Comput. Graph., 30(1):551–561, 2024. doi: 10.1109/TVCG.2023. 3326908 9

    2024 doi

  67. [82]

    T. Zhou, G. Y .-Y . Chan, S. Guo, J. Hoffswell, C. Xiao, V . S. Bursztyn, and E. Koh. Data Pictorial: Deconstructing raster images for data-aware animated vector posters. In Proc. UIST Adjunct, article no. 98, 3 pages. ACM, 2024. doi: 10.1145/3672539.3686353 2

    2024

  68. [83]

    L. Xie, Z. Zhou, K. Yu, Y . Wang, H. Qu, and S. Chen. Wakey-Wakey: Animate text by mimicking characters in a GIF. In Proc. UIST, article no. 98, 14 pages. ACM, 2023. doi: 10.1145/3586183.3606813 2

    2023

  69. [84]

    Zhu-Tian, Y

    C. Zhu-Tian, Y . Wang, Q. Wang, Y . Wang, and H. Qu. Towards automated infographic design: Deep learning-based auto-extraction of extensible timeline. IEEE Trans. Vis. Comput. Graph., 26(1):917–926, 2019. doi: 10. 1109/TVCG.2019.2934810 1, 2, 4 11

    2019

  70. [86]

    J. E. Zhang, N. Sultanum, A. Bezerianos, and F. Chevalier. DataQuilt: Extracting visual elements from images to craft pictorial visualizations. In Proc. CHI, article no. 45, 13 pages. ACM, 2020. doi: 10.1145/3313831. 3376172 1, 2

    2020 doi

  71. [90]

    T. Zhou, J. Huang, and G. Y .-Y . Chan. Epigraphics: Message-driven infographics authoring. In Proc. CHI, 18 pages. ACM, 2024. doi: 10. 1145/3613904.3642172 2

    2024

  72. [2016]

    doi: 10.1109/TVCG.2015.2467732 1

    2015

  73. [2017]

    This is the author’s version of the article that has been published in IEEE Transactions on Visualization and Computer Graphics

    doi: 10.1111/cgf.13193 2 10 © 2025 IEEE. This is the author’s version of the article that has been published in IEEE Transactions on Visualization and Computer Graphics. The final version of this record is available at: xx.xxxx/TVCG.201x.xxxxxxx

    2025 doi

  74. [2019]

    doi: 10.1109/TVCG.2018.2865158 2

    2018

  75. [2021]

    doi: 10.1145/3411764.3445356 2

  76. [2025]

    doi: 10.1145/3706598.3714155 2, 4, 8, 9

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.